OpenCV去除图像背景

去除图像背景，需要综合使用二值化（thresholding）、腐蚀（erosion）、膨胀（dilation）以及位运算（bitwise operations），代码如下：

#include "opencv2/imgcodecs.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/imgproc.hpp"
#include <iostream>

using namespace cv;

int main(int argc, char** argv)
{
    CommandLineParser parser(argc, argv, "{@input | dog.jpg | input image}");
    // Read an image
    Mat src = imread(samples::findFile(parser.get<String>("@input")));
    if (src.empty())
    {
        std::cout << "Could not open or find the image!\n" << std::endl;
        std::cout << "Usage: " << argv[0] << " <Input image>" << std::endl;
        return EXIT_FAILURE;
    }

    // Convert the image to grayscale
    Mat grayImg;
    cvtColor(src, grayImg, COLOR_BGR2GRAY);

    // Remove the background using a threshold
    // median filter is applied to reduce noise in the image
    // ksize is 5
    Mat grayImgBlurred;
    medianBlur(grayImg, grayImgBlurred, 5);

    // A binary threshold is applied to the grayscale image using a threshold
    Mat binaryImg;
    double thresh = threshold(grayImgBlurred, binaryImg, 150, 255, THRESH_BINARY_INV);

    // Output the thresh
    std::cout << thresh << std::endl;

    // The binary image is eroded to remove small objects and fill in small gaps using erode
    Mat erodedMask;
    erode(binaryImg, erodedMask, getStructuringElement(MORPH_RECT, Size(3, 3)), Point(-1, 1), 2);

    // The binary image is dilated to expand the remaining foreground objects # and fill in gaps using dilate
    Mat mask;
    dilate(erodedMask, mask, getStructuringElement(MORPH_RECT, Size(3, 3)), Point(-1, 1), 2);

    // The original input image is combined with the binary mask using bitwise_and
    Mat backgroundRemovedImg;
    bitwise_and(src, src, backgroundRemovedImg, mask);

    // Display the processed images
    imshow("Background Removed Image", backgroundRemovedImg);
    waitKey(0);

    return EXIT_SUCCESS;
}

###Background removal is removing the background from an image

import cv2

# Read an image
img = cv2.imread('../data/dog.jpg')
# Convert the image to grayscale
gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
  
# Remove the background using a threshold
# median filter is applied to reduce noise in the image
gray_img = cv2.medianBlur(gray_img, 5)

# A binary threshold is applied to the grayscale image using a threshold
ret, thresh = cv2.threshold(gray_img, 150, 255, cv2.THRESH_BINARY_INV)

# The binary image is eroded to remove small objects and fill in small gaps using erode
mask = cv2.erode(thresh, None, iterations=2)

# The binary image is dilated to expand the remaining foreground objects # and fill in gaps using dilate
mask = cv2.dilate(mask, None, iterations=2)
# The original input image is combined with the binary mask using bitwise_and
background_removed_img = cv2.bitwise_and(img, img, mask=mask)

# Display the processed images 
cv2.imshow('Background Removed Image', background_removed_img)


# Wait for a key press and then close the windows
cv2.waitKey(0)
cv2.destroyAllWindows()

原图：

处理后：

显然，这个结果并不美丽，我们可以尝试修改参数修正，但结果总是难以令人满意。于是采用更好更省事的办法，引用rembg库，调用u2net模型，去除背景。

使用rembg库去除图像背景

rembg库地址为：https://pypi.org/project/rembg/，这是一个基于机器学习模型的库，安装命令如下：

pip install rembg

如果有CUDA，可以安装GPU版：

pip install rembg[gpu]

使用该库，去除图像背景的代码如下：

### For background removal using rembg library

from rembg import remove
import cv2

#input path for image
input_path = '../data/dog.jpg'
output_path = 'output.png'

#read the image
input = cv2.imread(input_path)
output = remove(input)
# save the image 
cv2.imwrite(output_path, output)

# Display the processed images 
img = cv2.imread('output.png')
cv2.imshow('Background Removed Image', img)

# Wait for a key press and then close the windows
cv2.waitKey(0)
cv2.destroyAllWindows()

运行效果如下：

效果较之于第一种方法，更简洁，当然，安装的包也是很多的。